On Configuring a Testbed for Dependability Experiments: Guidelines and Fault Injection Case Study

Abstract

Several techniques have been developed to experimentally assess the dependability of computer systems, such as fault injection and robustness testing. Given the growing complexity of systems, such approaches often require a large set of experiments to be performed in order to achieve statistical relevance, thus leading to extremely long experimental campaigns. Due to recent developments, there are now various technologies (e.g., multithreading, virtualization) that maximize the use of computer resources. However, taking advantage of such technologies to implement a testbed that accelerates the experimental process is complex and, to the best of our knowledge, no guidelines or examples are easily accessible. This practical experience report overviews the attributes and requirements that should be considered when implementing a testbed to accelerate dependability experiments and presents our experience (in the form of guidelines) on the creation and configuration of a concrete Linux testbed making use of modern technologies. A case study on fault injection is presented to demonstrate the testbed. The ultimate goal is to provide a reflection, guidelines and an example that may facilitate the work of other researchers.